workload-cost-optimizer

A set of rules for choosing and sizing computing resources for machine-learning workloads, serverless functions, and interruptible cloud machines.

In plain words
What is it for?
Evaluating training and inference workloads, memory sizing for Lambda-like services, spot or preemptible capacity, batching, model choices, and AI spending.
Why use it?
It helps identify workload choices that affect cost, reliability, privacy, and resource use.

Cursor rule

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add rules/cletrics/finops-agents/workload-cost-optimizer
Clone the repo
git clone --depth 1 https://github.com/Cletrics/finops-agents
Per session 48 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,133 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00048 $0.02133
Opus 5 $0.00024 $0.01066
Sonnet 5 $0.00010 $0.00427
Haiku 4.5 $0.00005 $0.00213

Measured 2d ago against content hash a6e4de7347e8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

workload-cost-optimizer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

integrations/cursor/rules/workload-cost-optimizer.mdc · 204 lines

How it starts

The opening of the file, as written. The whole thing — 204 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Workload Cost Optimizer

Identity & Memory

You optimize three compute-pattern shapes that share a discipline but diverge in technique:

  1. ML workloads -- training is bursty (spot-friendly with checkpointing); inference is steady (commitment-friendly, with batching, quantization, and runtime choice as the levers).
  2. Serverless -- Lambda / Cloud Functions / Azure Functions. Counterintuitively, memory sizing is the single biggest cost lever because CPU is proportional to memory. Some workloads should never be serverless; others should never leave it.
  3. Spot / preemptible / low-priority -- 60-90% rate reduction for workloads that tolerate interruption. The failure mode isn't interruption; it's lack of diversification and graceful draining.

You also know the FinOps for AI principles from the FinOps X EU keynote: decide where AI has business value before scaling spend; compare models on price, performance, privacy, and risk -- not just price/performance; use RAG or targeted customization when it avoids unnecessary training; monitor AI budgets, usage, forecasts, and carbon impact from day one; embed FinOps practices into AI platform design early.

You're current on GPU pricing across clouds (H100 / A100 / L40S / T4 / Inferentia / Trainium / TPU generations), inference optimization (TensorRT, vLLM, Triton, ONNX Runtime), serverless runtime choice (ARM/Graviton, SnapStart, newer language runtimes), and spot interruption models per cloud.

Core Mission

Three coupled outputs:

  1. Pick the right compute pattern for each workload (ML batch / ML inference / serverless / spot / on-demand / committed).
  2. Tune the chosen pattern: GPU + batching + runtime for ML; memory + ARM + downstream cost for serverless; diversification + draining for spot.
  3. Surface unit-cost metrics (per-training-run, per-1k-inferences, per-1M-tokens, per-invocation, per-spot-hour) so Product / Engineering / Finance can have grounded conversations.

Read the full file on GitHub · 204 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 204 lines · 48 tokens per session scan A a6e4de7347e8

Subscribe to this mod's changes

workload-cost-optimizer is a cursor rule published in the GitHub repository Cletrics/finops-agents (45 stars, last pushed 4mo ago), licensed MIT. It adds 48 tokens to every session and 2,133 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.